China Issues First National Humanoid Robot Standards
China has released its first national standards for humanoid robots and embodied AI. The new regulations cover ethics, safety, and the full product lifecycle, from brain-like computing to deployment. This move establishes a key regulatory framework that will shape the development and commercialization of advanced robotics and physical AI agents in the country.
The new national standards were developed by a technical committee under the Ministry of Industry and Information Technology (MIIT), involving over 120 research institutions and companies. This body, the Humanoid Robots and Embodied Intelligence Standardization (HEIS) technical committee, was established in late 2025 to bring order to a rapidly growing sector. Key players from the Beijing Economic-Technological Development Area, including the Beijing Humanoid Robot Innovation Center (BHRIC), were major contributors, leading the drafting of technical requirements for motion control. The framework is extensive, breaking down standards into six core components: basic commonality, brain-like and intelligent computing, limbs and components, complete machines, application, and safety/ethics. The "brain-like" computing standards are particularly detailed, regulating the entire data lifecycle for model training and deployment. This move is designed to solve "industrial headaches" like production consistency and to establish clear evaluation metrics for robot intelligence levels. This standardization effort follows a period of explosive growth in 2025, which the MIIT has called the first year of mass production for humanoid robots in China. During that year, over 140 domestic manufacturers released more than 330 different models. The government has set a clear goal of establishing a comprehensive innovation system for these robots by 2025 and achieving mass production capabilities by 2027. The push for embodied AI is a distinct part of China's broader strategy to achieve Artificial General Intelligence (AGI), differing from the large language model-centric approach often seen in the U.S. Beijing is promoting a "pilot first, scale later" approach, encouraging local governments to experiment with different applications and development strategies. This has led to regional specializations, with Shanghai focusing on core hardware and Guangdong on complete humanoid platforms with companies like UBTECH and Unitree. In Beijing's tech ecosystem, startups like Manus are gaining prominence in the AI agent space. Manus has developed a multi-agent architecture that coordinates specialized sub-agents to handle complex workflows, a feature not yet standard in the industry. This aligns with a broader trend in China of adapting open-source global frameworks like LangChain and AutoGen to integrate with local large language models from companies such as Tencent and Baidu. Consumer adoption of AI is surging, providing a massive domestic market for AI agent applications. By the end of 2025, China's generative AI user base hit 602 million, with an adoption rate of 42.8%. A significant majority of these users—over 90%—express a preference for domestic AI models from companies like DeepSeek and Alibaba. This trend is shifting enterprise AI use from demand generation toward operational efficiency and cost reduction in areas like logistics and customer service.
